Anthropic: AI hasn’t caused job losses yet, but hiring is slowing
AI has yet to trigger the wave of job losses many fear, according to one of the sector’s biggest players, but early signs suggest it may already be reshaping how companies hire.
New research from Anthropic found little evidence that AI has increased unemployment in the occupations most exposed to the technology since the launch of ChatGPT in late 2022.
However, the study suggests entry into some highly exposed white-collar professions is beginning to slow, particularly for younger workers.
The report introduces a new measure of “observed exposure”, combining theoretical AI uses with real-world usage data from Anthropic’s chatbot, Claude.
The metric aims to capture not just what AI could do in theory, but what tasks it is actually being used for in professional settings.
The researchers found that large language models (LLMs) remain far from reaching their potential impact across the economy.
In computer and mathematical roles, among the sectors most associated with AI adoption, current AI usage covers around a third of job tasks, despite models theoretically being capable of assisting with far more.
Vulnerable jobs and sectors
The report ranks computer programmers, customer service representatives and data entry workers among the professions most exposed to AI tools.
Yet despite this high exposure, analysis of US labour market data found no statistically significant rise in unemployment among workers in those fields.
Instead, the earliest potential labour market signal appears in hiring trends.
Anthropic found the rate at which workers aged 22 to 25 are starting new jobs in highly exposed occupations has fallen compared with pre-AI levels.
The decline amounts to roughly a 14 per cent drop in the job-finding rate relative to 2022, although the researchers caution that the result is only marginally statistically significant.
The pattern broadly aligns with separate research suggesting that AI may be reducing demand for entry-level white-collar work while leaving experienced professionals less affected.
Older, female, and the highly-paid at risk
The study also highlights a notable demographic divide in AI exposure. Workers in the most exposed occupations are more likely to be older, female, highly educated and higher-paid than those in roles where AI currently plays little or no part.
Around 30 per cent of occupations show no meaningful AI coverage at all, including many hands-on jobs such as cooks, mechanics, bartenders and lifeguards.
More broadly, the research suggests that while AI has the potential to automate or accelerate a wide range of tasks, adoption in the real economy remains gradual.
The report says the gap between theoretical capability and real-world deployment may persist for years, as companies navigate legal constraints, technical integration challenges and the need for human oversight.
That cautious pace contrasts with the intensity of public debate about AI and employment.
Recent corporate decisions, including large-scale layoffs at tech companies shifting towards AI-driven operations, have only heightened fears that automation could rapidly displace workers.
But for now, Anthropic’s analysis suggests the labour market impact of generative AI remains subtle.